Elon Musk’s AI venture, xAI, has failed to meet its own deadline for releasing a comprehensive AI safety framework, as highlighted by the watchdog group The Midas Project. This follows concerns about xAI’s commitment to AI safety, which has come under scrutiny for its AI chatbot, Grok, which has been reported to make inappropriate requests, like undressing photos of women, and exhibit unfiltered language compared to competitors like Gemini and ChatGPT.
In February, during the AI Seoul Summit, xAI presented a draft outlining its approach to AI safety, detailing safety priorities and deployment considerations. However, The Midas Project pointed out that this draft only addressed future AI models that are not currently under development, lacking concrete strategies for identifying and mitigating risks—a critical aspect of a safety agreement the company had previously signed.
xAI initially promised to revise its safety policy within three months, targeting a release by May 10, but this deadline went unacknowledged on any of the company’s official platforms.
Despite Musk’s warnings about the risks associated with unchecked AI development, xAI continues to have a poor safety record. According to a recent study by SaferAI, a nonprofit focused on enhancing accountability in AI laboratories, xAI is rated poorly among its competitors due to its ineffective risk management practices.
It’s important to note that other AI companies, such as Google and OpenAI, have not performed significantly better in terms of safety. Many have expedited their safety testing and either delayed or completely omitted the publication of safety reports. This trend raises concerns as AI technologies advance, highlighting an alarming prioritisation of speed over safety at a time when emerging AI capabilities could pose greater risks.
In summary, xAI’s failure to deliver on its safety commitments underscores broader challenges within the AI industry regarding maintaining safety amidst rapid technological progress.
Fanpage: TechArena.au
Watch more about AI – Artificial Intelligence
